Hierarchical switch fault diagnosis based on transformer algorithm in four-leg inverters of stand-alone wind energy conversion systems

With the increasing development of renewable energy resources, stand-alone structures are gaining more attention. Among these, wind energy systems are particularly notable because of their advantages, including sustainability, low operational expenses, and minimal environmental impact. Due to the challenges of load balancing in such systems, four-leg inverters have emerged as a viable solution, offering improved performance under unbalanced load conditions. However, like all inverters, they remain susceptible to internal faults. Accordingly, this paper proposes a hierarchical two-level Transformer-based model to detect switch internal faults, including open-circuit and short-circuit in four-leg inverters. The OPAL-RT hardware-in-the-loop setup was used to generate data in various scenarios to validate the efficiency of the proposed framework. The results demonstrate that the developed technique can effectively classify fault types and identify faulty switches compared to state-of-the-art algorithms and single-level structures. • A wind energy conversion system with a four-leg inverter and standard controls is modeled to assess how DC-link and neutral-leg dynamics affect fault detection. • Both open-circuit and short-circuit switch faults are examined across wide operating ranges, including balanced and unbalanced loads. • A two-level hierarchical Transformer is proposed, outperforming single-level schemes and other state-of-the-art machine learning algorithms.